Foundation models for radiology—the position of the AI for Health Imaging (AI4HI) network

Abstract Foundation models are large models trained on big data which can be used for downstream tasks. In radiology, these models can potentially address several gaps in fairness and generalization, as they can be trained on massive datasets without labelled data and adapted to tasks requiring data...

Full description

Saved in:
Bibliographic Details
Main Authors: José Guilherme de Almeida, Leonor Cerdá Alberich, Gianna Tsakou, Kostas Marias, Manolis Tsiknakis, Karim Lekadir, Luis Marti-Bonmati, Nikolaos Papanikolaou
Format: Article
Language:English
Published: SpringerOpen 2025-08-01
Series:Insights into Imaging
Subjects:
Online Access:https://doi.org/10.1186/s13244-025-02056-9
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849343911182467072
author José Guilherme de Almeida
Leonor Cerdá Alberich
Gianna Tsakou
Kostas Marias
Manolis Tsiknakis
Karim Lekadir
Luis Marti-Bonmati
Nikolaos Papanikolaou
author_facet José Guilherme de Almeida
Leonor Cerdá Alberich
Gianna Tsakou
Kostas Marias
Manolis Tsiknakis
Karim Lekadir
Luis Marti-Bonmati
Nikolaos Papanikolaou
author_sort José Guilherme de Almeida
collection DOAJ
description Abstract Foundation models are large models trained on big data which can be used for downstream tasks. In radiology, these models can potentially address several gaps in fairness and generalization, as they can be trained on massive datasets without labelled data and adapted to tasks requiring data with a small number of descriptions. This reduces one of the limiting bottlenecks in clinical model construction—data annotation—as these models can be trained through a variety of techniques that require little more than radiological images with or without their corresponding radiological reports. However, foundation models may be insufficient as they are affected—to a smaller extent when compared with traditional supervised learning approaches—by the same issues that lead to underperforming models, such as a lack of transparency/explainability, and biases. To address these issues, we advocate that the development of foundation models should not only be pursued but also accompanied by the development of a decentralized clinical validation and continuous training framework. This does not guarantee the resolution of the problems associated with foundation models, but it enables developers, clinicians and patients to know when, how and why models should be updated, creating a clinical AI ecosystem that is better capable of serving all stakeholders. Critical relevance statement Foundation models may mitigate issues like bias and poor generalization in radiology AI, but challenges persist. We propose a decentralized, cross-institutional framework for continuous validation and training to enhance model reliability, safety, and clinical utility. Key Points Foundation models trained on large datasets reduce annotation burdens and improve fairness and generalization in radiology. Despite improvements, they still face challenges like limited transparency, explainability, and residual biases. A decentralized, cross-institutional framework for clinical validation and continuous training can strengthen reliability and inclusivity in clinical AI. Graphical Abstract
format Article
id doaj-art-c9409718e8154de6be2850bcf80fb5a8
institution Kabale University
issn 1869-4101
language English
publishDate 2025-08-01
publisher SpringerOpen
record_format Article
series Insights into Imaging
spelling doaj-art-c9409718e8154de6be2850bcf80fb5a82025-08-20T03:42:49ZengSpringerOpenInsights into Imaging1869-41012025-08-0116111410.1186/s13244-025-02056-9Foundation models for radiology—the position of the AI for Health Imaging (AI4HI) networkJosé Guilherme de Almeida0Leonor Cerdá Alberich1Gianna Tsakou2Kostas Marias3Manolis Tsiknakis4Karim Lekadir5Luis Marti-Bonmati6Nikolaos Papanikolaou7Champalimaud FoundationLa Fe Health Research InstituteR&D Department of Maggioli SpA, Greek BranchHellenic Mediterranean University and Foundation for Research and Technology HellasHellenic Mediterranean University and Foundation for Research and Technology HellasUniversitat de Barcelona, Artificial Intelligence in Medicine Lab (BCN-AIM), Department of Mathematics and Computer ScienceLa Fe Health Research InstituteChampalimaud FoundationAbstract Foundation models are large models trained on big data which can be used for downstream tasks. In radiology, these models can potentially address several gaps in fairness and generalization, as they can be trained on massive datasets without labelled data and adapted to tasks requiring data with a small number of descriptions. This reduces one of the limiting bottlenecks in clinical model construction—data annotation—as these models can be trained through a variety of techniques that require little more than radiological images with or without their corresponding radiological reports. However, foundation models may be insufficient as they are affected—to a smaller extent when compared with traditional supervised learning approaches—by the same issues that lead to underperforming models, such as a lack of transparency/explainability, and biases. To address these issues, we advocate that the development of foundation models should not only be pursued but also accompanied by the development of a decentralized clinical validation and continuous training framework. This does not guarantee the resolution of the problems associated with foundation models, but it enables developers, clinicians and patients to know when, how and why models should be updated, creating a clinical AI ecosystem that is better capable of serving all stakeholders. Critical relevance statement Foundation models may mitigate issues like bias and poor generalization in radiology AI, but challenges persist. We propose a decentralized, cross-institutional framework for continuous validation and training to enhance model reliability, safety, and clinical utility. Key Points Foundation models trained on large datasets reduce annotation burdens and improve fairness and generalization in radiology. Despite improvements, they still face challenges like limited transparency, explainability, and residual biases. A decentralized, cross-institutional framework for clinical validation and continuous training can strengthen reliability and inclusivity in clinical AI. Graphical Abstracthttps://doi.org/10.1186/s13244-025-02056-9Artificial intelligenceRadiologyFoundation modelsBias
spellingShingle José Guilherme de Almeida
Leonor Cerdá Alberich
Gianna Tsakou
Kostas Marias
Manolis Tsiknakis
Karim Lekadir
Luis Marti-Bonmati
Nikolaos Papanikolaou
Foundation models for radiology—the position of the AI for Health Imaging (AI4HI) network
Insights into Imaging
Artificial intelligence
Radiology
Foundation models
Bias
title Foundation models for radiology—the position of the AI for Health Imaging (AI4HI) network
title_full Foundation models for radiology—the position of the AI for Health Imaging (AI4HI) network
title_fullStr Foundation models for radiology—the position of the AI for Health Imaging (AI4HI) network
title_full_unstemmed Foundation models for radiology—the position of the AI for Health Imaging (AI4HI) network
title_short Foundation models for radiology—the position of the AI for Health Imaging (AI4HI) network
title_sort foundation models for radiology the position of the ai for health imaging ai4hi network
topic Artificial intelligence
Radiology
Foundation models
Bias
url https://doi.org/10.1186/s13244-025-02056-9
work_keys_str_mv AT joseguilhermedealmeida foundationmodelsforradiologythepositionoftheaiforhealthimagingai4hinetwork
AT leonorcerdaalberich foundationmodelsforradiologythepositionoftheaiforhealthimagingai4hinetwork
AT giannatsakou foundationmodelsforradiologythepositionoftheaiforhealthimagingai4hinetwork
AT kostasmarias foundationmodelsforradiologythepositionoftheaiforhealthimagingai4hinetwork
AT manolistsiknakis foundationmodelsforradiologythepositionoftheaiforhealthimagingai4hinetwork
AT karimlekadir foundationmodelsforradiologythepositionoftheaiforhealthimagingai4hinetwork
AT luismartibonmati foundationmodelsforradiologythepositionoftheaiforhealthimagingai4hinetwork
AT nikolaospapanikolaou foundationmodelsforradiologythepositionoftheaiforhealthimagingai4hinetwork